Convert every Microsoft Sentinel analytic rule in a solution into a Defender XDR Custom Detection YAML under the solution's XDR Detections folder.
78%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
Fix and improve this skill with Tessl
tessl review fix ./.github/skills/sentinel-xdr-rule-converter/SKILL.mdRun the CLI from the repository root:
python -m sentinel_xdr_migration.cli <command>If the package is not installed, first run:
python -m pip install -e Tools\SentinelToXDRMigrationSolutions/.Run sentinel-xdr-migration doctor.
If firstRun is true, run
sentinel-xdr-migration setup --non-interactive. Continue offline when
authentication is unavailable.
Run:
sentinel-xdr-migration inspect --solution "<solution-path>"Review the analytic-rule count and source paths.
If deterministic rewrites are needed, create
Solutions/<solution>/Logs/sentinel-xdr-migration/<run-id>/migration-config.yaml.
Run:
sentinel-xdr-migration convert --solution "<solution-path>"Do not overwrite conflicting generated files without explicit approval.
Report:
Solutions/<solution>/Logs/sentinel-xdr-migration/<run-id>/transformation-report.html
path.Do not stop a batch or ask for a per-rule decision solely because a Sentinel
entity identifier has no exact Defender XDR mapping. Select the best supported
mapping from the source identifiers and columns actually emitted by the query,
preserve the original value as an alert customDetails output-column binding,
and mark the assumption for human review in conversion provenance. For example,
an available Host.FullName may be provisionally mapped to the supported
hosts.nameColumn; keep the original output column in custom details and
explicitly state that this is an assumption, not proven host identity or parity.
Prefer a directly supported matching identifier when present; never invent a
projected column, unsupported Graph field, or data transformation.
Best-effort entity review is nonblocking for conversion, structural validation, and read-only runtime query/schema tests, so large batches can continue without per-rule pauses. It remains visibly pending and hash-bound to the source and converted query. It is not an acceptance or review decision and does not clear other findings. Missing/projected-away values, conflicting bindings, custom detail limits, malformed content, or other technical/KQL failures retain their normal errors and gates. Do not package or deploy content with unresolved review flags; query execution and preserved custom details do not prove entity correlation or alert parity.
The converter must not modify Package/mainTemplate.json, and must not modify Analytic Rules except to bump the top-level patch version when an already-tracked Analytic Rule has been modified and --overwrite is used.
Generated detections must remain disabled.
XDR Detections must contain detection YAML only.
New detections receive an independent top-level version (defaulting to the solution's release version in [3.1.0, 4.0.0) or 3.1.0); source
Sentinel versions are preserved in contentProvenance.source.version.
The XDR release must be a major.minor.patch string in [3.1.0, 4.0.0).
When reconverting with --overwrite, unmodified detections preserve their existing version, while modified Analytic Rules and modified Custom Detections automatically bump their patch version (X.Y.Z -> X.Y.(Z+1)). Invalid versions or mismatched source identities
are conflicts and must not be silently reset. Older YAML without its own version
fails validation/packaging; after reviewing manual edits and obtaining overwrite
approval, reconvert with --overwrite to initialize 3.1.0, or explicitly author
the independent XDR version. Do not change schema/API versions to satisfy the XDR release bounds.
Azure/Azure-Sentinel
c8a3ab1
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.